Immersive audio for desktop systems

Abstract

Numerous applications are envisioned for integrated media workstations to create, edit, and accurately monitor digital video and audio. There are two major classes of limitations that impede the performance of current desktop loudspeaker-based sound systems. The first encompasses problems that arise due to the local acoustical environment, such as early reflections from the CRT and nearby flat surfaces, as well as the design characteristics and performance of the loudspeakers. The second class involves limitations that arise from variations in human listening characteristics and listener movement relative to the loudspeakers. In this paper several findings are presented that are based on acoustical, psychoacoustical, and signal processing methods for delivering accurate sound field representations. Furthermore, a novel vision-based method for accurate listener tracking is presented that eliminates the need for headgear or tethered magnetic devices. Current findings show that this algorithm can be implemented without imposing a significant computational overhead to the host processor. Research directions include adaptive real-time HRTF synthesis based on the tracking information, as well as vision-based pinna shape classification for improved performance.